team-memory

skill
Guvenlik Denetimi
Uyari
Health Uyari
  • License — License: MIT
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Low visibility — Only 5 GitHub stars
Code Uyari
  • network request — Outbound network request in frontend/src/app.js
  • network request — Outbound network request in frontend/src/pages/setup.js
Permissions Gecti
  • Permissions — No dangerous permissions requested

Bu listing icin henuz AI raporu yok.

SUMMARY

AI memory for teams, git version controlled with web UI and CLI skills

README.md

team-memory social preview

team-memory

Shared, version-controlled knowledge base your whole team builds together

Locally run in your CLI or UI, backed up to your git knowledge repo automatically

License: MIT


Most team knowledge lives in people's heads, scattered Slack threads, or stale wikis nobody trusts. team-memory is different: it lives in a Git repo your team already owns, every entry is a commit, and anyone can open the web UI and jot something down in seconds.

Two ways to build it up:

  • Web UI — lightweight note-taking interface at http://team-mem. Capture a decision, a gotcha, a workflow pattern — whenever it crosses your mind, not just at the end of a session
  • Claude Code hook — fires automatically when a session ends, summarises what was built, and commits a structured entry hands-free

team-memory web UI


Features

Auto-save sessions Hook fires when Claude Code ends, summarises with Claude, commits structured entry to your repo
Web UI Browse, save, lookup, and flag stale entries at http://127.0.0.1:7438
Smart categorisation LLM picks the right file from your index — or skip it and save directly
GitHub-backed storage Plain Markdown in a repo you own — edit, search, and version-control it directly
MCP server Claude Code can read and write your knowledge base mid-session
No telemetry Binds 127.0.0.1 only, never phones home

Install

macOS / Linux

curl -LsSf https://raw.githubusercontent.com/AndrewSkea/team-memory/master/install.sh | sh

Windows (PowerShell)

irm https://raw.githubusercontent.com/AndrewSkea/team-memory/master/install.ps1 | iex

The installer downloads the binary, adds it to ~/bin (or %USERPROFILE%\bin on Windows), prompts for your GitHub PAT and memory repo, wires Stop/PreCompact hooks into ~/.claude/settings.json, and registers the MCP server in ~/.claude.json. Config is written with mode 600 on Unix.

Prerequisites: a GitHub repo and a fine-grained PAT with contents:write on that repo.

What needs elevated privileges? On Linux/macOS, install.sh only calls sudo if you pick service mode — to append 127.0.0.1 team-mem to /etc/hosts and (Linux) setcap cap_net_bind_service on the binary so it can bind port 80. Skip service mode to avoid sudo entirely. On Windows, install.ps1 requires no admin by default; running it elevated additionally configures netsh portproxy and the hosts entry.

Binary install location trust: the binary lives under your user profile (~/bin or %USERPROFILE%\bin). Anything that can write there can replace the binary — keep that directory's ACL/permissions as they default for your user. For a multi-user machine, install to a system path manually and run from there instead.

Service mode (always-on web UI)

The installer offers two run modes. Service mode keeps the web UI available at all times:

Non-admin (default) Admin (Run as Administrator)
Autostart Registry HKCU\...\Run key Task Scheduler (RunLevel Highest)
Browser URL http://127.0.0.1:7438/ http://team-mem/
team-mem hostname Manual (see below) Auto-added to hosts file
Port 80 redirect netsh interface portproxy 80 → 7438

To get the http://team-mem/ shortcut without re-running as admin, add one line to C:\Windows\System32\drivers\etc\hosts (requires admin once):

127.0.0.1 team-mem

Then access via http://team-mem:7438/ (no portproxy without admin).


Usage

After install, use Claude Code normally. When a session ends, team-memory-mcp automatically summarises it and commits a structured entry to your memory repo. Open http://127.0.0.1:7438 to browse your knowledge base, save entries manually, or search by keyword.

Model & cost

All LLM calls (categorisation, summary, reminders) go through the claude CLI pinned to Haiku (--model haiku). Haiku is roughly an order of magnitude cheaper than Sonnet/Opus and is more than adequate for these short, structured generations. Override per-process with the TEAM_MEMORY_MODEL env var if you ever want a bigger model:

TEAM_MEMORY_MODEL=sonnet team-memory-mcp

Privacy & Security

  • PAT and Anthropic key stored in ~/.config/team-memory/config.json (CLI) and browser localStorage (web UI)
  • Binary binds 127.0.0.1 only — never exposed to the network
  • No telemetry, no third-party scripts

See SECURITY.md for the full threat model.


Contributing

See CONTRIBUTING.md for build instructions, project layout, and code style.

make build   # build binary locally
make test    # run all tests

MIT License · © Andrew Skea

Yorumlar (0)

Sonuc bulunamadi